Time: 09:00-10:30, September 22, 2022

Title: Language Understanding in Task-oriented Dialogue Systems

Abstract : Task-based dialogue systems have received extensive attention from industry and academia. Among them, dialogue language understanding (DLU) is the core component, which has developed rapidly in recent years. This report first reviews and summarizes the development of the DLU field in recent years, especially the methods in fewer-labeled scenarios, and then gives the future development trend of the DLU field.

Speaker: Wanxiang Che, a professor and doctoral supervisor of Computing Faculty at Harbin Institute of Technology, Deputy Dean of the Institute of Artificial Intelligence, and Deputy Director of the Research Center for Social Computing and Information Retrieval. Heilongjiang Province "Longjiang Scholar" young scholar, visiting scholar at Stanford University. He is currently a deputy director and secretary-general of the Professional Committee of Computational Linguistics of the Chinese Information Society of China; executive committee and secretary-general of the Asia-Pacific Chapter of the Association for Computational Linguistics (AACL); a senior member of the China Computer Federation. He has published more than 200 academic papers in high-level journals and conferences at home and abroad, such as ACL, EMNLP, AAAI, and IJCAI, among which AAAI's 2013 paper won the Outstanding Paper Honorable Mention Award. He has published 4 textbooks and 2 translated books. At present, he is undertaking a number of scientific research projects such as the 2030 "New Generation Artificial Intelligence" major project and the National Natural Science Foundation of China. He has won many awards such as the 2020 Heilongjiang Youth Science and Technology Award.

Time: 10:45-12:15, September 22, 2022

Title: Cross-modal Knowledge Acquisition and Fusion

Abstract : Text, image, and audio signals carry rich knowledge of the world and human experience. Due to their inherent differences in various aspects, knowledge does not distribute uniformly across these signals. Acquiring and fusing knowledge from multi-modal signals to improve both the quantity and quality of knowledge accessible by models has attracted extensive attention in recent years. In this talk, we will summarize the progress of cross-modal knowledge acquisition and fusion, hoping to highlight the trends in this field. And we will also discuss the future directions worth pursuing in this field.

Speaker: Peng Li is a Research Associate Professor at Institute for AI Industry Research (AIR), Tsinghua University. His main research interests include natural language processing, pre-trained models, cross-modal information processing, question answering, information extraction, machine translation, and dialogue system. He has published more than 60 papers in international conferences and journals on artificial intelligence and won the first place on a few highly influential leaderboards. He served as the Area Chair of COLING 2022 and the Senior Area Chair of AACL 2022. He was the recipient of First Prize of Qian Weichang Chinese Information Processing Science and Technology Award and Second Prize of Chinese Institute of Electronics Science and Technology Progress Award.

Time: 14:00-15:30, September 22, 2022

Title: Affective Computing for Recommender Systems

Abstract : Product reviews contain a wealth of semantic information that provides a basis for us to enhance personalized recommendation. The powerful semantic extraction and comprehension capacity of deep neural network can overcome the defects of the bag-of-word model in traditional text semantic analysis, help us better portray the user's preferences and product characteristics, and also provide material for the interpretability of recommendations through affective computing. This tutorial will introduce the review-based recommendation techniques in recent years, and will mainly focuse the design choices and the interpretability of recommendation, and finally look discuss the potential trend of review-based recommendation systems.

Speaker: Chenliang Li is a full professor at School of Cyber Science and Engineering, Wuhan University. His research interests inlcude information retrieval, natural language processing and social media analysis. He has published over 70 research papers on leading academic conferences and journals such as SIGIR, ACL, WWW, IJCAI, AAAI, TKDE and TOIS. He has served as Associate Editor / Editorial Board Member for ACM TOIS, ACM TALLIP, IPM and JASIST. His research won the SIGIR 2016 Best Student Paper Honorable Mention and TKDE Featured Spotlight Paper. He is a recipient of ACM Wuhan Rising Star Award.

Time: 15:45-17:15, September 22, 2022

Title: Quantum Uncertainty in Natural language

Abstract : In recent years, the cross research of quantum theory and artificial intelligence has attracted more and more attention. This lecture will explain the fundamental connection between natural language and quantum mechanics, and introduce the frontier progress of quantum natural language processing and future challenges for the key scientific problems of quantum uncertainty in natural language.

Speaker: Peng Zhang is a professor and doctoral supervisor, vice Dean of the School of Computer Science of Tianjin University. He has devoted himself to the research work of quantum information retrieval and quantum language modeling for more than ten years, and actively promoted the application of the quantum language model. He published top conference papers including NeurIPS, SIGIR, ICLR, ACL, IJCAI, AAAI, WWW, CIKM, EMNLP and leading journal papers such as TNNLS, TKDE, TIST, IP&M. He has received SIGIR 2017 Best Paper Award Honorable Mention and ECIR 2011 Best Poster Award.

Time: 18:45-20:15, September 22, 2022

Title: Conversational Information Seeking and Recommendation

Abstract : In contrast to traditional information retrieval and recommendation systems, conversational information seeking and conversational recommender systems aim to solve search and recommendation tasks through conversational interactions between the system and the user. As a new paradigm for interacting with search engines, conversational information seeking can better capture user intent through conversational interactions, and users can get answers directly without having to search results, in response to the inflexibility of traditional retrieval for complex intent queries. Conversational recommender systems can reveal users' current preferences more quickly through real-time, multi-round online conversational interactions and better understand the reasons behind consumer behaviour. In recent years, research on conversational information seeking and conversational recommender systems has received increasing attention and has been developing rapidly. This tutorial will summarize recent studies on conversational information seeking techniques and conversational recommendation from different perspectives, and present the progress of research on conversational information seeking and conversational recommender systems in terms of methods, data, evaluation metrics, and the challenges.

Speaker: Zhaochun Ren a professor at Shandong University. His research interests lie in information retrieval and natural language processing, with emphasis on conversational information retrieval, recommender systems, social media analysis, dialogue systems, opinion mining, and summarization. I aim to develop intelligent agents that can address complex user requests and solve core challenges in IR and NLP towards that goal. His research has appeared at various prestigious conferences and journals, including SIGIR, WWW, ACL, KDD, WSDM, CIKM, AAAI, IJCAI, EMNLP, ACM TOIS, IEEE TKDE, IPM, etc.. He has received the Best Full Paper Runner Up award in CIKM 2017, and the Best Student Paper award in WSDM 2018.

Time: 20:30-22:00, September 22, 2022

Title: Automatic Solving Math Problem in Neuro-symbolic Computing

Abstract : Automatically solving mathematical problems is a fundamental challenge in neural symbolic computing, which is divided into: (1) Formal mathematical problems rely on pre-built formal systems, requiring models to perform multi-step reasoning and interact with the formal system to complete problem solving or theorem proving. (2) Informal math problems do not depend on formal systems. Given a mathematical problem described in natural language, the model needs to complete the reasoning process and return the final answer by itself. Recently pre-trained language models have been successfully applied to formal/informal math problems, but SOTA results are still below expert level. There are still many unanswered questions on this topic. In this tutorial, we will first introduce the definition and background of mathematical problems, then take two classic problems, automatic theorem proving and math word problems as examples, review the technical development, and finally discuss the future direction of the topic.

Speaker: Yichun Yin received his Ph.D. from Peking University in 2018. Currently working as a senior researcher in Speech and Semantics Lab of Huawei's Noah's Ark, he is mainly engaged in the research of efficient pre-trained language models and neuro-symbolic computing. He published many papers at natural language processing conferences such as ACL and EMNLP, one of which is the most cited paper in EMNLP2020.

Time: 9:00-10:30, September 23, 2022

Title: Towards Automated Fact Checking

Abstract : Automated Fact Checking aims to verify a given claim according to retrieved evidence, and has received increasing attention in recent years. The key components to solve this task is to collect sufficient relevant evidence according to the given claim from multiple sources, digest the evidence, and finally make the veracity prediction. The community has made promising progress by addressing multiple evidence formats, the fusion of evidence pieces, model explainability and robustness. In this tutorial, we will provide a gentle introduction to automated fact checking, including the task setups, datasets, models, and applications. We will focus on the most recent advance, including effective evidence selection and organization, fusion of structured and unstructured evidence pieces, and how to improve model robustness in real-world scenarios.

Speaker: Dr. Yansong Feng is an associate professor in the Wangxuan Institute of Computer Technology at Peking University. Before that, he obtained his PhD from ICCS (now ILCC) at the University of Edinburgh. His current research interests include using probabilistic methods to distill knowledge from large volumes of natural language texts, and supporting intelligent human-computer interfaces, such as question answering and dialogue. He has served as Action Editor and Area Chair for ARR and *ACL conferences. Yansong received the IBM Faculty Award in 2014 and 2015,  and the IBM Global Shared University Research Award in 2016.

Time: 10:45-12:15, September 23, 2022

Title: k-nearest-neighbor machine translation

Abstract : Abstract:One important change for machine translation in the deep learning era is that the translation knowledge are no longer represented in a symbolic way but embedded in the parameters of the neural networks. However, even large scale neural networks cannot learn all the knowledge in the training data, especially for the low frequency  events. k-nearest-neighbor machine translation is a retrieval based technique. kNNMT employs a translation datastore with symbolic translation knowledge to assist neural machine translation models, showing great potential in modeling low-frequency events, fast adaptation, etc. This talk covers both the basis of kNN-MT and its recent advances, including dynamically integrate the symbolic knowledge into the neural system, how to control the size of the symbolic knowledge base and the interpretability of this framework.

Speaker: Shujian Huang is currently an associate professor and phd advisor in Nanjing University. His research interests includes machine translation, text analysis and understanding, etc. He is now the dupty director of the Techinical Committee on Machine Translation of CIPSC, and senior member of CCF. He currently serves as AC/AE/SPC for major NLP/AI conferences such as ACL, EMNLP, NAACL, AAAI, IJCAI, etc.